Sequential Subset Selection Procedure of Random Subset Size for Early Phase Clinical Trials

Abstract

In early phase clinical trials, the objective is often to select a subset of promising candidate treatments whose treatment effects are greater than the remaining candidates by at least a pre-specified amount to bring forward for phase III confirmatory testing. Under certain constraints such as budgetary limitations or difficulty of recruitment, a procedure which selects a subset of fixed, pre-specified size is entirely appropriate, especially when the number of treatments available for further testing is limited. However, clinicians and researchers often demand to identify all efficacious treatments in the screening process and a subset selection of fixed size may not be sufficient to satisfy the requirement as the number of efficacious treatments is unknown prior to the experiment. To address this issue, we discuss a family of sequential subset selection procedures which identify a subset of efficacious treatments of random size, thereby avoiding the need to pre-specify the subset size. Various versions of the procedure allow adaptive sequential elimination of inferior treatments and sequential recruitment of superior treatments as the experiment processes. We compare these new procedures with Gupta’s random subset size procedure for selecting the one best candidate by simulation.